Before this past long weekend I was testing with Windows and getting
about a 1% difference; but I didn't report it here because I was using
the 'wallclock' time and not CPU time, on Windows. (that was with
Even's 200 map draw test) I personally feel that the difference is so
minimal now that yes, we should enable SUPPORTS=POINT_Z_M for the next
MS4W release.
-jeff
--
Jeff McKenna
MapServer Consulting and Training Services
http://www.gatewaygeomatics.com/
On 2017-07-03 4:49 AM, Daniel Hardes wrote:
Do we need more testing or could "USE_POINT_Z_M = ON" the default in the
next release?
Even Rouault <[email protected]> hat am 30. Juni 2017 um
16:40 geschrieben:
On vendredi 30 juin 2017 09:59:30 CEST Daniel Morissette wrote:
> On 2017-06-30 9:50 AM, Even Rouault wrote:
> > I'm wondering if we shouldn't revisit the decision to have
USE_POINT_Z_M
> > disabled by default. This is often annoying.
> >
> > Apparently, adding USE_POINT_Z_M defaulting to OFF was driven by
> > performance reasons per
> > https://github.com/mapserver/mapserver/issues/1244 . But this was 12
> > years ago. Perhaps with recent hardware the performance difference
isn't
> > that big. The ticket unfortunately doesn't contain a test case to
bench
> > (likely lots of lines/polygons, or large lines/polygons to render). If
> > the perf difference is still there, a more involved change would be
> > indeed to move the z and m components into separate arrays.
>
> #1244 refers to #1224 which is using the gmap demo mapfile... which is a
> relatively small dataset, so it should be easy to verify if the issue is
> still relevant on more recent hardware using gmap or probably any other
> dataset.
>
> https://github.com/mapserver/mapserver-import/issues/1224
OK, I downloaded
http://dl.maptools.org/dl/gmap-ms46.tar.gz
(not sure if there's a more uptodate version of it).
Added the necessary MAP and SYMBOLSET keywords in the mapfile and
symbolset file.
With master, compiled in Release (-O3)
and the following Python mapscript bench.py
{{{
import mapscript
for i in range(200):
map = mapscript.mapObj('gmap75.map')
img1 = map.draw()
}}}
I ran it 10 times consecutively with each config and took the smallest
time ("real" figure)
With USE_POINT_Z_M=OFF
time python bench.py
real 0m4.167s
user 0m3.964s
sys 0m0.200s
With USE_POINT_Z_M=ON
time python bench.py
real 0m4.252s
user 0m3.972s
sys 0m0.260s
So USE_POINT_Z_M=ON is 2% slower than OFF
On a recent (~ 1 year old) laptop with a
Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz
Even
--
Spatialys - Geospatial professional services
http://www.spatialys.com
_______________________________________________
mapserver-users mailing list
[email protected]
https://lists.osgeo.org/mailman/listinfo/mapserver-users
